This paper proposes a full-frame video stabilization algorithm based on an improved bit-plane matching and image mosaicing.

A image stabilizing algorithm based on a fast locally refined gray-coded bit-plane matching based on diamond search and median-based motion correlation which greatly reduces the computational load while maintaining matching precision.

The image motion is obtained by the gray project and then the motion data is accurately detected by the grid detection and bit plane match method.

An equivalent error coding algorithm defined in the paper is also presented which reduces the number of encoded bit-planes and decrease the final output data rate once again.

In the light of the human visual system (HVS) and the strong correlativity of coefficients among the sub-images, a quantizer with different step-size is applied to each coefficient, and then by simple and efficient frame bit-plane structure, the coefficients are scanned and compressively encoded.

Through saving row-status in advance to eliminate the correlation between strips,encoding the logic contexts based on single window and computing the bit-plane status with priority,the algorithm can achieve the parallel encoding of bit-plane and three passes in the bit-plane by a single scanning.

Denoising method of images based on their decomposed bit planes;

Then the paper proposes two kinds of technologies for wavelet transforming coding, one is a coding technology based on zerotree, the other is a coding technology based on information source partition and bit plane.

The experimental results show that our method achieves higher PSNR and better subjective effects than other methods even when fewer bit planes are transmitted.