Transforming Pixels: Crafting a 3D Integer Discrete Cosine Transform for Advanced Image Compression

We propose an innovative technique for image compression based on the 3-dimensional Integer Discrete Cosine Transform (3D-Integer DCT), which will serve as an alternative to the existing DCT-based compression technique. If an image is encoded as cubes [row x column x temporal length] instead of blocks [row x column], higher compression can be achieved. Here, the number of blocks is represented as the temporal length. To construct cubes, we use highly correlated blocks, and the correlation level is determined using the mean absolute difference (MAD). The suggested 3D-Integer DCT-based coder can achieve a higher compression ratio while maintaining the required image quality. It also needs fewer coefficients to encode an image than the usual Joint Photographic Expert Group (JPEG) coder. Adopting integer DCT further reduces the computational complexity of the proposed algorithm, given the abundance of methods available in the literature to determine equivalent integers for DCT. We choose an optimum integer group that minimizes mean squared error (MSE) and improves coding efficiency for computing 3D-Integer DCT. We also conducted a detailed analysis to examine the impact of implementing integer DCT in image compression. When we look at peak signal-to-noise noise ratio (PSNR), bits per pixel, and structural similarity index (SSIM), we see that the proposed algorithm does a better job than the standard real-value DCT-based compression algorithm like JPEG.