Page 96 - The-5th-MCAIT2021-eProceeding

America’s National Cancer Institute’s (NCI) Genomic Data Commons (GDC) portal. A total of 85 pathological
images were randomly selected from the GDC portal to form the dataset for this research.

2.2. Algorithm testing

For this research, four algorithms have been selected for testing. The four algorithms selected for this
research are:
● Discrete Cosine Transform (DCT), a lossy image compression algorithm (Maru 2020)
● Discrete Wavelet Transform (DWT), a lossy image compression algorithm (Nain 2020)
● Huffman Encoding, a lossless image compression algorithm (Kai-Meng 2020)
● Set Partitioning in Hierarchical Trees (SPIHT), a near-lossless image compression algorithm (Fangfang
2020)

These four algorithms were selected for this research as the first three algorithms represent algorithms that
being widely used in image compression of other types of image (Mantoro 2017). By comparing the SPIHT
algorithm against these algorithms, it would be a fair head-to-head comparison and representation of how these
algorithms would perform in a daily use case (Rahman 2019).

2.3. Performance Evaluation

There can be numerous ways to evaluate a compression algorithm, whether it be through compression speed,
a measure of how fast an algorithm processes an image and compresses it, through compression ratio, a measure
of how much more compact the compressed image is when compared to the original file in size, through image
quality, a measure of how similar the reconstructed image is when compared to the original image (Hussain
2018). These measures serve as a quantifiable measurement to evaluate the efficiency of an image compression
algorithm.
In order to measure the compression speed of an algorithm, encoding time is used, which refers to the amount
of time taken, in seconds, for the encoder of a compression algorithm to encode and compress the data in the
original input image into the outputted compressed data (Sharma 2017).
Compression Ratio (CR) refers to the ratio of the number of bits between the original uncompressed file and
the compressed file. It is used to measure how much space has been compressed by the algorithm in the image.
For example, a CR of 4:1 signifies that the compressed image is only about 1/4th of the size of the original
image (Rahman 2019). The equation to determine CR is as illustrated as follows:

( ) = (1)

Although both are used to quantify the amount of compression that is applied to a file, CR is not to be
confused with compression rate, measured in Bits Per Pixel (BPP), which represents the number of bits in
average, that is needed to represent a pixel in an image (Hussain 2018). The equation to measure BPP is as
illustrated as follows:

( ) = (2)

There are multiple ways to measure image quality when attempting to evaluate an image compression
algorithm. Commonly used methods to measure image quality would be Peak Signal-to-Noise Ratio (PSNR)
and Structural Similarity Index Measure (SSIM). PSNR is measured in decibels (dB) and represents the quality
of an image relative to the size of the error, where a high PSNR value represents a low measure of error in the
reconstructed image when compared to the original (Hussain 2018). The equation to measure PSNR is as
illustrated as follows, where n represents the number of bits that represent the pixel:

E- Proceedings of The 5th International Multi-Conference on Artificial Intelligence Technology (MCAIT 2021) [83]
Artificial Intelligence in the 4th Industrial Revolution