Adaptive transform coding of video signals

In the paper, five schemes for the adaptive transform coding of video signals are presented. Adaptation is based on adaptive quantisation and adaptive bit selection. A Max quantiser having a Laplacian density distribution is used to achieve adaptive quantisation, whereas classification according to the activity within the transform block and based on the human visual characteristics are used for adaptive bit selection. The discrete cosine transform and the Hadamard transform are used to transform three source pictures of different statistics, and results are compared to find the best scheme for each transform and picture. Photographic results are presented to see the effects of the distortions introduced by different transforms on different source pictures due to data compression. Owing to the limitations of the photographic process, the subjective quality of the pictures as perceived from the video monitor cannot be accurately represented. However they are useful for comparison purposes.