Over the past few decades, digital image compression has been a hot area of research, and with the development of new media storing devices, fresh image compression schemes have arise. Among the digital data processed nowadays, images have become very popular, for their characteristic of being able to represent content information in a vast number of approaches.
In this study a lossless compression method to compress Bayer images using mixed-lattice transforms is presented. With help of the Lifting scheme and its perfect reconstruction characteristics, a lossless compression method is able to be performed.
The transform coefficients are entropy coded with a Golomb-Rice coder that models well the probability distributions of these coefficients. The Golomb-Rice coder is made adaptive through the update of parameter k that determines the length of the codewords. The adaptation is done using local statistics of previously encoded data.
By separating the image in three different Bayer components and adapting a second generation wavelet, higher compression ratios than JPEG-LS were obtained. The algorithm is simple enough to achieve an average of twenty one computations per pixel, and because it has a memory less characteristic it can be employed in systems with low storage capacity.
