Image Compression Using DCT with Graceful Quality Degradation

There are two kinds of image compression: lossless and lossy. In both kinds, the objective is to reduce the required space to represent the image. In lossless image compression, there is no loss of information, so the image quality is not altered. However, in lossy image compression methods, some of the information is lost, and the quality of the image is affected. The challenge with lossy image compression is to select the most irrelevant information for our visual system and compromise it to gain compression. The lost information affects the image quality, but the visual quality does not suffer much. This is called graceful degradation of the image quality.

The image compression algorithm implemented in this project is lossy. We’ll take a block of size $8\times8$ from a grayscale image and transform it into the frequency domain to select the information we want to discard or compromise. Every possible block is processed in the same way—in a serial manner. In this project, we’ll also see the extent to which the visual quality degrades as we compromise more and more information to achieve a higher compression ratio.

The task of producing compressed image data from a given image is achieved by an encoder. The decoder is responsible for reconstructing the image from the compressed data. The efficacy and compromise on quality is measured by comparing the original and reconstructed images.

The following diagram illustrates how the encoder and decoder interact: