Colour image compression algorithm based on the DCT transform using difference lookup table

Abstract

A simple and efficient method for lossy colour image compression is proposed. Here, the discrete cosine transform (DCT) is applied to the YCbCr image obtained from the original RGB image. The bisection method is used to define the required threshold for a prefixed user peak signal to noise ratio as a controlled quality criterion. The thresholded and quantised DCT coefficients are encoded with a new technique. The proposed technique uses the difference of the indexes of the retained coefficients in coordination with DCT block adaptive scanning to encode efficiently the coefficients. The difference of the indexes is stored in a lookup table called (dLUT). When compared with recent methods, the obtained results show that the proposed algorithm achieves high performance. Introduction: The compression can lead to an efficient transmission and storage of images. It reduces the size of the images by decreasing the redundancy that contains. In general, there are two kind of compression: lossless and lossy compression. The lossless compression guaranties that the original and reconstructed images are identical. The efficiency and the compression ratios are weak compared with those of the lossy compression. In this last, the quality of the reconstructed image is degraded to an acceptable level in the benefit of high compression ratios. The lossy compression of the colour image has been widely studied and it can be further categorised in two principle methods [1]: † The direct methods act directly on the image samples as in block truncation [2] and vector quantisation based techniques [3]. † The transform methods use transforms to concentrate the energy of the image in a few number of coefficients. Many transforms are used such as principal component analysis [4], discrete cosine transform (DCT) [5] and wavelets [6].

Saliency enabled compression in JPEG framework

“…The value of PSNR (in dB) shown in Table 3 for proposed method, and JPEG baseline is for the most important regions of the image and for rest of the methods for the whole image. When any DWT or DCT methods in [6, 10–12, 14] are applied on the most important regions , the mean‐square error (MSE) of these regions is higher than overall MSE. The reason for this is while applying these transform‐based methods in a region with high variance, the energy compaction is lesser compared to a region with lower variance [18], which results in higher MSE after quantisation to achieve lower bit‐rate.…”

“…On the other hand, the proposed method gives a stable curve for the most important regions as well as for the overall image. Table 3 presents the performance comparison of the proposed method with the recently published DCT-and DWT-based algorithms in [6,11,12,14] and the JPEG baseline [10] on the same test images as shown in Fig. 8.…”

“…The experimental results obtained on different colour images clearly showed that our proposed method outperforms the recently published similar methods in terms of the perceptual quality of the reconstructed images. By ranking 8 × 8 blocks, we were able to reconstruct the ROI at the decoder side more accurately than the Table 3 Performance comparison between the proposed method, JPEG baseline [10], colour block truncation code -pattern fitting (CBTC-PF) [6], colour image compression by DCT with adaptive block scanning (CDABS) [11], genetic algorithm-discrete wavelet transform (GA-DWT) [14], and difference lookup table (dLUT) [12] algorithms Image R 1 area, % JPEG [10] CBTC-PF [6] CDABS [11] GA-DWT [14] dLUT [ recent state-of-the-art works where the ROI is approximated by a rectangular bounding box. The average overhead is also found to get reduced by 36.38% by using delta encoding as a postprocessing on rank information matrix.…”

“…The methods under the second category are called transform methods [1], where the image is transformed into frequency domain. Principle component analysis [9], discrete cosine transform (DCT) [6,[10][11][12], and discrete wavelet transform (DWT) [13][14][15] are the most popular transformations used for this purpose. Such transformations concentrate the energy of the image in a few number of coefficients, making it suitable for removing perceptual redundancies.…”

Adaptive Lossy Image Compression Based on Singular Value Decomposition

“…Compression techniques [7]- [16] play an important role in the storage and transmission of image data. Their main purpose is to represent an image in a very compact way, that is, through a small number of bits without losing the essential content of the information present in the image.…”