International Journal of Engineering and Technology

Transmission of uncompressed video segments requires more bandwidth and need more storage for video data. Earlier approaches produces compressed videos , where there exists data loss and they produce less effective results at decompressing video segment. We propose a new lossless approach to perform compression of video segment, where snapshot level video compression is done by splitting the video into tiny snapshots. The snapshots are trained with the neural network and the trained values are used to perform video compression. The input video is de-framed to generate sequence of scenes and compressed with the help of wavelet transform. The efficiency of wavelet helps to reduce the signals into small set of signals. For each frame generated from the video, the features of the scene are extracted and used to identify the frames of scenes. A feature variance matrix is generated by identifying the variance of features between subsequent frames of any scene identified. Based on computed variance matrix, the unaffected pixels are neutralized and the pixels with variance are kept original. Then the frames are transformed with wavelet, and the transformed signals are applied with neural network to come up with output signals. The proposed method has produced higher rate of compression and produces efficient results in decompression. Also the proposed method overcomes the problem of loss introduced in video compression by other methods.