Exploring Pixel Intensity for Wavelet Transform Compression Methods: An

Analytical Study

H. A. Abdulkareem, M. D. Almustapha, H. Adamu, U. F. Abdu-Aguye, H. Bello and I. K. Musa

Department of Electronics & Telecommunications Engineering; 3Centre for Energy Research & Training,
Ahmadu Bello University, Zaria, Nigeria Corresponding
Author: [email protected]
Abstract: This paper focuses on the significance of image and video compression in the context of
multimedia advancements and widespread utilization of graphical images in mobile networks. The main
objective of compression techniques is to decrease the size of images and videos while maintaining their
quality, thereby facilitating efficient transmission and storage. In this study, we analyze the utilization of
wavelet transforms and pixel intensity for compression. Six video samples, including four acquired ones
and two benchmark samples, were employed to implement the proposed technique. To address the
potential quality degradation in the compressed images, a luminance enhancement model was applied.
The simulation results demonstrate the effectiveness of the proposed method, indicating improved
contrast pixel intensity. Specifically, the Enhanced Lifting Wavelet Transform (E-LWT) compression
technique achieved the highest Peak Signal-to-Noise Ratio (PSNR) values across all six video samples,
followed by the Enhanced Discrete Wavelet Transform (E-DWT). For individual video frames
(NAERLS1.avi, NAERLS2.avi, NTA1.avi, and NTA2.avi), the E-LWT technique exhibited percentage
improvements of 12.59%, 5.10%, 4.71%, and 1.93% over E-DWT. Moreover, for benchmark video
frames (Akiyo.avi and Forman.avi), the E-LWT technique also demonstrated percentage improvements of
14.46% and 5.31% over E-DWT.

Index Terms: Image Compression, luminance pixel intensity, PSNR

I. INTRODUCTION
Usually, images are compressed immediately after capture to decrease data size while preserving image
quality for human viewing, ignoring the editing needed for computer visualization (Liu et al., 2022; Sahib
et al., 2022; Walaa et al., 2020; Zainab et al., 2022). Image processing encompasses various fields such as
Image Segmentation, Image Compression, and Image Enhancement. Image Compression is a component
of Image processing where images and videos undergo compression (Harish and Richa, 2011;
Mohammed and Anas, 2022). Two types of wavelets exist: continuous wavelet transform and discrete
wavelet transform (Prabhjot, 2015; Abdulkareem1 et al., 2018). Standard methods for compressing still
images (e.g., JPEG) and motion pictures (e.g., MPEG) are based on the DCT (Abdulkareem3 et al., 2018;
Meera et al., 2019; Prasannajit, 2022), but they have limitations at high compression ratios (Frank, 2010).
At low data rates, DCT-based transforms suffer from a "blocking effect" and other drawbacks like
mosquito noise and aliasing distortions (Frank, 2010). To overcome these limitations, the Discrete
Wavelet Transform (DWT) has gained importance (Adil and Kamil, 2019). The DWT provides space-
frequency decomposition, enabling energy compaction at low-frequency subbands and edge localization
at high-frequency subbands (Frank, 2010; BOSE et al., 2022). DWT is based on a wavelet function that
satisfies multi-resolution analysis requirements (Christian et al., 2009; Adil and Kamil, 2019). DWT
represents images at different resolution levels and converts them into high-pass and low-pass wavelet
coefficient series. This transformation is applied recursively on the low-pass series until the desired
number of iterations is reached (Priyanka et al., 2011). The lifting scheme is introduced for efficient DWT
computation, reducing computation time and simplifying the process (Chesta et al., 2011). It involves
three operations: split, predict, and update.

2. METHODOLOGY

1
The following are the highlighted steps in this research, which involve analysing the compression
technique of wavelet transforms using pixel intensity in a step-by-step manner.
i. Video Acquisition
ii. Implementation of the DCT
iii. Implementation of the DWT
iv. Implementation of the LWT
v. Application of the developed Brightness Enhancement

2.1 Video Acquisition

To efficiently evaluate the performance of the proposed techniques under various conditions, a total of
four acquired video data and two benchmark video data were utilized. The video frames sample, as
presented in (Abdulkareem1 et al., 2018), was derived from these videos. The initial four sample videos,
namely NAERLS1.avi, NAERLS2.avi, NTA1.avi, and NTA2.avi, were captured using a video camera.
The remaining two videos, serving as benchmarks, were sourced from an online internet image
processing database. The video information was obtained using the Matrix Laboratory (MATLAB
R2015a) image processing toolbox command, and the details are provided in Table 1.
Table 1 Sample of Video Data

Collected Data
Sample of
File Name
Video Data File File Frames
Size(*.avi)
1 NAERLS1.avi 18.1Mb 157

2 NAERLS2.avi 10.3Mb 155

3 NTA1.avi 9.6Mb 152
4 NTA2.avi 11.2Mb 200

5 Akiyo.avi 11Mb 300

6 Foreman.avi 7.25Mb 100

For easy processing and analysis Table 1, were initially converted into frame of static images.
2.2 Implementation of the Discrete Cosine Transform (DCT)
To effectively assess the performance of the proposed techniques across diverse conditions, a
combination of four acquired video datasets and two benchmark video datasets were employed.
The video frames sample, as outlined in (Abdulkareem1 et al., 2018), was extracted from these
videos. The initial set of four sample videos (NAERLS1.avi, NAERLS2.avi, NTA1.avi, and
NTA2.avi) were captured using a video camera, while the remaining two videos, which acted as
benchmarks, were sourced from an online image processing database. The video information was
acquired using the image processing toolbox command of

2
 Matrix Laboratory (MATLABR2015a),
Start

Input Video
Data

Existing Method
Extract the Frames
from the Video Data

Divide Frames into

NN
Blocks

No
Size appropriate Resize

Yes

Encode the Frames

and perform DCT

Apply Inverse
Achieved Proposed

Transform
Enhancement

Apply the
Enhancement

Output Compressed
result

Stop

Figure 1: Flowchart of MATLAB Implementation of Improved DCT Video Image

From Figure 1, the first hidden block represents the existing DCT method while the second hidden block
represents the modification achieved improvement.

2.3 Implementation of the DWT

At every level of the wavelet transform (Li and Drew, 2003), four output images (approximate, vertical,
horizontal, and diagonal) details are obtained. In this research, the 2D wavelet transform was
implemented by multiplying the wavelet function by the scaling function as follows (Li and Drew, 2003).

(1)
After which the four details of the image compression were determined as follows (Li and Drew,
2003):
Approximate detail
(2)
Horizontal detail
(3)
Vertical Detail
(4)
Diagonal Detail
(5)
The approximate detail is repeatedly passed through a low pass (L) and a high pass (H) filter bank
until an appropriate level of compression is achieved as (Li and Drew, 2003):

3
 (6)

Start

Input Video

Existing M ethod
Data

Extract the Frames

from the Video Data

Divide Frames into

NN
Blocks

Size appropriate Resiz e

No
Yes

Encode the Frames

and perform DWT Achieved Proposed

Apply Inverse
Transform
Enhancement

Apply the
Enhancement

Output Compresse d
result

Stop

Figure 2: MATLAB Implementation of Improved DWT Video Image.

Figure 2 illustrates the implementation of Improved DWT Video Image through a flowchart. Initially, the
video is imported and read within the MATLAB simulation environment. The total number of frames is
determined, followed by the conversion of the video into individual frames. These frames are
subsequently extracted and divided into segmented blocks of size N x N. A decision box evaluates
whether the segmented blocks are of appropriate sizes. If the answer is "No," the blocks undergo resizing.
If the answer is "Yes," the frames are encoded and subjected to the DWT process. Finally, the compressed
image is obtained by applying the inverse transform.

2.4 Implementation of the LWT

The Lifting Wavelet Transform (LWT) is a transform method designed to overcome certain difficulties
encountered in the Discrete Wavelet Transform (DWT), while maintaining computational efficiency. As a
result, the implementation of LWT is similar to DWT, but with the same number of samples at each stage
as the initial set of samples. In this study, the input image sample obtained from the processed sample
video was divided into two sets of samples (even and odd) to enable the efficient lifting filter, ensuring
accurate approximation and extraction of details. The step-by-step procedural approach for implementing
the improved LWT-based compression is outlined as follows: Input the sampled enhanced frames, that is,

4
 where, and j = 1, 2, …M+1. The three stages (split, prediction, and update) of
,
LWT on the image were then performed as:
i. The input image signal was split into even, and odd, samples as follows:
(1)
(2)
ii. The integer positions of the odd samples from the neighbouring even samples were predicted as
follows:
(3)
where, is the resulting prediction residuals or high sub-band coefficients.
Assuming the sample pixels have a strong correlation in the angle and the integer pixels are marked
by , the half pixels by , and the quarter pixels by . The prediction of is taken
as a linear combination of the even samples as follows:
(4)
where,

(5)

The weighting factor is given by the filter coefficients.

iii. In the updating step, the even samples are replaced using the following equation:
(6)

The values of are always located at an integer position which is one of the characteristics of the
LWT.

The LWT saves a significant number of memories as the samples to be stored align with the input for
each stage. In addition, the number of computations needed is decreased as approximation coefficients
can be obtained from previously computed details of input samples. The implementation of the LWT
follows a similar process to DWT and DCT, with the compression technique being applied only during
the encoding stage. Figure 3 represents the flowchart implementation of Improved LWT Video Image,
where the video is read from the initial block after starting.

5
Figure 3: Flowchart of MATLAB Implementation of Improved LWT Video Image

Application of the developed Brightness Enhancement

Although the standard colour transformation is relatively easy to implement and many researchers have
actually adopted it for coloured video frames, it still has great challenges, especially when the objects in a
particular video frame have very similar grey scale values. This research used a different enhancement
technique as in (Abdulkareem1 et al., 2018)

3. RESULTS AND DISCUSSION

The performance of the enhanced compression technique is evaluated. Since the essence of compression
is to reduce the size of the video data for easy transmission, the performance of the enhanced compression
technique is evaluated using sample size (bytes), compression ratio, and peak signal-to-noise ratio (dB).
The standard DCT, DWT, and LWT were developed, implemented, and their resultant compressed
outputs were enhanced using the developed brightness enhancement model.

6
Table 1: Simulation Result of Performance Comparison of the Sample Size after Compression
Sample Original Size DWT DCT LWT
NAERLS1.avi 18.1Mb 12.38Mb 7.10Mb 6.41Mb
NAERLS2.avi 10.3Mb 8.74Mb 6.93Mb 5.30Mb
NTA1.avi 9.60Mb 7.36Mb 5.43Mb 5.06Mb
NTA2.avi 11.2Mb 9.22Mb 7.60Mb 7.12Mb
Akiyo.avi 11.0Mb 3.45Mb 2.81Mb 2.11Mb
Forman.avi 7.25Mb 0.205Mb 0.165Mb 0.196Mb

Table 1 shows a noticeable reduction in size through compression techniques, prior to applying brightness
enhancement methods. Additionally, histogram distribution improves pixel intensity for enhanced image
representation. Table 2 presents the compression ratio analysis results of the different techniques after
applying the brightness enhancement mode.

Table 2: Simulation Result of Compression Ratio Analysis

Sample DWT DCT LWT FOA
NAERLS1.avi 13.698 15.201 22.983 26.356
NAERLS2.avi 11.146 11.204 11.656 14.811
NTA1.avi 10.939 11.127 17.901 16.222
NTA2.avi 11.493 12.841 14.710 20.100
Akiyo.avi 10.620 11.541 18.424 21.001
Forman.avi 11.475 11.868 14.081 14.750
Increasing the compression ratio leads to a decline in image quality due to artifacts caused by the block-
based scheme. This indicates that a higher compression ratio signifies a greater reduction in signal due to
compression. Table 2 demonstrates that the LWT technique achieves superior compression compared to
DCT and DWT for the NAERLS1.avi sample video. The LWT technique shows compression ratio
improvements of 33.86% and 40.40% over DCT and DWT, respectively. Additionally, the DCT
technique exhibits a compression ratio improvement of 9.89% over DWT. Similarly, for the
NAERLS2.avi sample video, the LWT technique outperforms DCT and DWT with improvements of
3.88% and 4.38%, respectively, while the DCT technique provides a 0.52% improvement over DWT.

For the NTA sampled videos, the LWT technique achieves compression improvements of 37.84% and
38.89% over DCT and DWT, respectively, for the NTA1 sample. Likewise, for the NTA2 sample, the
LWT technique yields improvements of 12.70% and 21.87% over DCT and DWT, respectively. In both
NTA samples, DCT performs better than DWT, with improvements of 1.69% and 10.50% for NTA1 and
NTA2, respectively.

When considering the benchmark video frames, the LWT technique also achieves superior compression,
with improvements of 37.36% and 42.36% over DCT and DWT, respectively, for the Akiyo.avi
benchmark video. Similarly, for the Forman.avi benchmark video, the LWT technique provides
improvements of 15.72% and 18.51% over DCT and DWT, respectively. DCT performs better than DWT
with improvements of 7.98% and 3.31% for Akiyo.avi and Forman.avi benchmark videos, respectively.
In all the techniques considered, LWT-based compression consistently produces the best results compared
to DCT and DWT techniques. Table 3 evaluates the Peak Signal to Noise Ratio (PSNR) for various
techniques used after reconstructing the compressed output.

7
 Table 3: Simulation Results of Peak Signal-to-Noise Ratio (PSNR) of Various Techniques before and
after their Enhancement
Sample DWT E_DWT % DCT E_DCT % LWT E_LWT %
NAERLS1.avi 19.23dB 20.42dB 5.83 18.98dB 20.43dB 7.10 21.89dB 23.36dB 6.30
NAERLS2.avi 15.75dB 16.93dB 6.97 16.76dB 17.35dB 3.40 16.92dB 17.84dB 5.16
NTA1.avi 15.09dB 15.78dB 4.37 15.41dB 16.21dB 4.94 16.01dB 16.57dB 3.38
NTA2.avi 16.17dB 17.29dB 6.48 16.61dB 17.25dB 3.71 16.94dB 17.63dB 3.91
Akiyo.avi 17.40dB 18.04dB 3.55 17.54dB 18.28dB 4.05 20.17dB 21.09dB 4.00
Forman.avi 17.55dB 18.92dB 3.44 17.97dB 19.10dB 5.92 18.64dB 19.98dB 6.71

Table 3 shows result of peak signal-to-noise ratio (PSNR) of various techniques and after their
enhancement. From the Table, all the three techniques (DCT, DWT and LWT where significantly
improved when enhanced. Example, with the NAELS 1avi sample video DWT was enhanced by 5.83%,
DCT enhanced by 7.10%, LWT enhanced by 6.30%.

3.1 Comparison of Techniques before Enhancement

Table 4 displays the percentage improvement of LWT compared to DCT and DWT in terms of PSNR.
The results from Table 4 demonstrate that the LWT compression technique consistently achieved the
highest PSNR values across all six video samples, surpassing both DCT and DWT. In the case of
NAERLS1.avi, NAERLS2.avi, NTA1.avi, and NTA2.avi, the LWT yielded PSNR improvements of
13.65%, 1.18%, 3.75%, and 1.95% over DCT, respectively. Similarly, for the benchmark video frames
Akiyo.avi and Forman.avi, the LWT provided improvements of 13.04% and 3.61% over DCT.
Additionally, the LWT yielded PSNR improvements of 12.51%, 6.91%, 5.75%, and 4.55% over DWT for
NAERLS1.avi, NAERLS2.avi, NTA1.avi, and NTA2.avi, respectively. Similarly, for the benchmark
video frames Akiyo.avi and Forman.avi, the LWT achieved improvements of 13.04% and 5.85% over
DWT.

Table 4: Percentage Improvement of LWT over DCT and DWT for PSNR
Sample LWT DCT DWT Percentage (%) Percentage (%) Improvement
(dB) (dB) (dB) Improvement of LWT over DWT
of LWT over DCT
NAERLS1.avi 21.98 18.98 19.23 13.65 12.51
NAERLS2.avi 16.92 16.76 15.75 1.18 6.91
NTA1.avi 16.01 15.41 15.09 3.75 5.75
NTA2.avi 16.94 16.61 16.17 1.95 4.55
Akiyo.avi 20.17 17.54 17.40 13.04 13.04
Forman.avi 18.64 17.97 17.55 3.61 5.85

3.2 Comparison of Techniques after Enhancement

Table 5 shows percentage improvement of enhanced E_LWT over enhanced DCT and DWT for PSNR.
From the table, it has been observed that the E-LWT compression technique produced the highest PSNR
values in all the six video samples compared to the E-DWT and E-DCT For the respective individual
sample video frames of NAERLS1.avi, NAERLS2.avi, NTA1.avi, and NTA2.avi, the E_LWT produced

8
PSNR percentage improvement of 12.54%, 2.75%, 2.11% and 2.16% over E-DCT. However, for the
benchmark video frame of Akiyo.avi and Forman.avi, the E_LWT also produced a percentage
improvement of 13.32% and 4.40% over E-DCT. Also, for the respective individual sample video frames
of NAERLS1.avi, NAERLS2.avi, NTA1.avi, and NTA2.avi, the E_LWT produced PSNR percentage
improvement of 12.59%, 5.10%, 4.71% and 1.93% over E-DWT and for the benchmark video frame of
Akiyo.avi and Forman.avi, the E_LWT also produced a percentage improvement of 14.46% and
5.31% over E-DWT.

Table 5: Percentage Improvement of Enhanced LWT over Enhanced DCT and DWT for PSNR
Sample E_LWT E_DCT E_DWT Percentage (%) Percentage (%)
(dB) (dB) (dB) Improvement Improvement
of E_LWT of E_LWT
over E_DCT over E_DWT

NAERLS1.avi 23.36 20.43 20.42 12.54 12.59

NAERLS2.avi 17.84 17.35 16.93 2.75 5.10
4. CONCLUSION
This research analyzes the use of wavelet transform compression technique with pixel intensity.
Enhanced standard transform techniques were applied to six video samples (four acquired and two
benchmark) to improve image quality. The performance of the enhancement model was evaluated on
sampled video frames and then applied to enhance the output of the compression techniques. Simulation
results demonstrated the efficiency of the developed enhancement method, with improved pixel intensity
and histogram distribution in all video frames (Table 2). Peak signal-to-noise ratio (PSNR) evaluation
(Table 3) indicated that the enhanced techniques achieved better signal quality compared to the standards.
The analysis revealed that the E-LWT compression technique consistently produced the highest PSNR
values in all six video samples, outperforming E-DWT and E-DCT for the individual video frames
(NAERLS1.avi, NAERLS2.avi, NTA1.avi, NTA2.avi). Similarly, for the benchmark video frames
(Akiyo.avi and Forman.avi), E-LWT outperformed E-DWT and E-DCT.
REFERENCES
Adil Amirjanov1, Kamil Dimililer2 (2019) “Image compression system with an optimization of
compression ratio” IET Image Processing, Vol. 13 Iss. 11, pp. 1960-1969 © The Institution of
Engineering and Technology

BOSE A. LUNGISANI, CASPAR K. LEBEKWE, ADAMU MURTALA ZUNGERU, ABID YAHYA

(2022) “Image Compression Techniques in Wireless Sensor Networks: A Survey and
Comparison” IEEE Access, VOLUME 10, 2022, p 82515- 82530

Chris Solomon & Toby Breckon, (2010) “Fundamentals of Digital Image

Processing A Practical Approach with Examples in Matlab” School of Physical Sciences, Univers
ity of Kent, Canterbury, UK. vol. 5, pp 90-101. Retrieved on 25th December, 2014.

Frank Y. Shih (2010) “Image Processing and Pattern Recognition Fundamental and Techniques” Wiley I
EEE Press. 445 Hoes Lane Piscataway, NJ 08854. Retrieved on 3rd December, 2014.
H. A. ABDULKAREEM1, A. M. S. TEKANYI, I. YAU, K. A. ABU-BILAL, H. ADAMU (2018)
“Brightness Enhancement Technique for Video Frame Improvement Based on Pixel Intensity
Analysis” i-manager’s Journal on Image Processing, Vol. 5, No. 4. p1-8, ISSN: 2349 - 4530

H. A. ABDULKAREEM2, A. M. S. TEKANYI, I. YAU, K. A. ABU-BILAL, H. ADAMU (2018)

“Optimized Video Compression Using Modified Intelligent Behavior of Firefly Algorithm” i-
manager’s Journal on Image Processing, Vol. 5, No. 4. p 1-8, ISSN: 2349 - 4530

9
H. A. ABDULKAREEM3, A. M. S. TEKANYI, I. YAU AND H. A. ADAMU (2018) “An Efficient
Video Compression Method using Enhanced Discrete Cosine Transform Technique”
International Journal of Pure and Applied Sciences, VOL. 5 NO.1 (IJPAS), ISSN: 1660-5332.
P1-8
Harish Rohil, Richa Kukreja (2011) “Image Compression –A Comprehensive Study” International
Journal of Advanced Research in Computer Science, Volume 2, No. 4, p269-272

Walaa M. Abd-Elhafiez1, Wajeb Gharibi2, Mohamed Heshmat3 (2020) “An efficient color image
compression technique” TELKOMNIKA Telecommunication, Computing, Electronics and
Control, Vol. 18, No. 4, pp. 2371~2377

Ian T. Young, Jan J. Gerbrands, Lucas J. vanVliet (1995) “Fundamentals of

Image Processing” Delft University of Technology. Retrieved on 10th January, 2013.http://
www.google.com.ng/url?
sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CBwQFjAA&url=http%3A%2F
%2Frepository.tudelft.nl%2Fassets%2Fuuid%3A1d58e4e5 4a0365a0506808fcf2de82%2FImage
ProcessingFundamentals.pdf&ei=TKyTVcO6NLDQ7AbP_ruYAw&usg=AFQjCNGnGJjf29x3Q
mLsv_TJ 3N

Keshika Jangde, Mr. Rohit Raja, (2014) “Image Compression Based on Discrete Wavelet and Lifting
Wavelet Transform Technique” International Journal of Science, Engineering and Technology
Research (IJSETR), Volume 3, Issue 3, March 2014, p394-399
Linfeng Liu, Tong Chen, Haojie Liu , Shiliang Pu, Li Wang, Qiu Shen (2022) “2C-Net: Integrate Image
Compression and Classi” Research square Springer Nature 2021 LATEX template, pg2-26

Meera Thapar Khanna, Chetan Ralekar, Anurika Goel, Santanu Chaudhury1, Brejesh Lall (2019)
“Memorability-based image compression” IET Image Process, Vol. 13, Iss. 9, pp. 1490-1501
Mohammed K. Al-Obaidi 1 and Anas Fouad Ahmed (2022) “Implementation of image compression
based on singular value decomposition” Global Journal of Engineering and Technology
Advances, 11(03), p086–092, elSSN:2582-5003, https//gjeta.com/

Prabhjot kour (2015) “IMAGE PROCESSING USING DISCRETE WAVELET TRANSFORM” IPASJ I
nternational Journal of Electronics & Communication (IIJEC) Web Site:
Volume 3, Issue 1, January 2015 ISSN 2321-5984, Volume 3, p53-59
http://www.ipasj.org/IIJEC/IIJEC.htm

Prasannajit Dash (2022) “REVIEW ON MULTI OBJECTIVE OPTIMIZATION TECHNIQUES USED

FOR IMAGECOMPRESSION” IJRET: International Journal of Research in Engineering and
Technology, eISSN: 2319-1163, Volume: 04 Issue: 02, pISSN: 2321-7308, p1-8

Priyanka Singh, Priti Singh, Rakesh Kumar Sharma,(2011) “JPEG Image Compression based on
Biorthogonal, Coiflets and Daubechies Wavelet Families”, Applications (0975 – 8887), Volume
13– No.1. International Journal of Computer

Tejas S. Patel1, Ravindra Modi2, Keyur J. Patel3 (2013) “Image Compression

Using DWT and Vector Quantization” International Journal of Innovative
Research in Computer and Communication Engineering, Vol. 1, Issue 3, May 2013, p651-659

Zainab Jawad Ahmed1, Loay Edwar George2, Raad Ahmed Hadi3 (2022) “Images Compression using
Combined Scheme of Transform Coding” Special Issue on Multidisciplinary Sciences and

10
Advanced Technology, Journal of Engineering Research and Sciences, www.jenrs.com 1(9): p08-
14,

11