Directional Lifting-Based Wavelet Transform For Image Compression

The conventional discrete wavelet transform is based on a separable transform by using two one-dimensional discrete in the horizontal and vertical direction, which fails to provide an efficient representation.As we know, the natural images are all rich in texture imformation. Such a separable transform can't effectively utilize image texture or boundary information. It will have some influence in the image compressing and image denoising. The directional lifting wavelet transform suits far better to the image orientation features, while still have wonderful reconstruction. In order to obtain the best predictiing and updating signals, we must need to do the prediction in all candidate directions for all images to decide which direction is the best for us. Meanwhile, the Computation for one transform is large.This article based on the general ADL transform, proposes one new kind of adaptive directional lifting wavelet transform algorithm. The image is partitioned into many non-overlapping blocks using the algorithm. For each block the lifting wavelet is adaptively selected based on the image statistics (GLCM's angle second moment or Directioanl Filter Bank). The proposed GLCM algorithm and Directioanl Filter Bank method detect all the image blocks of image to decide whether the image block is homogeneous or not, before wavelet transform. The DFB method can judge the directional of the block further. To the direction information few blocks, uses ordinary horizontal and the vertical promotion directly; To direction information many blocks, uses the directional lifting wavelet.The experiments discussed the time used for transform and the image compression performance influence of the directional lifting transformation based on two different block method(GLCM and DFB method).We campare the Lossy compression results by using our method, ADL method and traditional method. Experimental results show that the transform time of the new method based on the two different partition method less than the transform time of ADL transform.In comparison, the compression performance of two methods and ADL method are better than traditional lifting wavelet transform. Meanwhile, the improved two directional lifting wavelet method and the ADL method compression performance varies slightly.