Medical image sequences motion estimation
|School||University of Electronic Science and Technology|
|Keywords||Wavelet Transform Quantum Genetic Algorithm Motion Estimation Block matching Medical image sequences|
Medical image sequence compression telemedicine system is an important technology. Compressed video sequences motion estimation is a key technology in its purpose is to remove the video sequence time correlation of adjacent frames. Block-matching motion estimation algorithm, the first prediction image into a plurality of the subject do not overlap each block, and a search range in a reference image search, according to a given block of the matching criteria to find the best match for each block , to obtain a motion vector of each block. Existing block matching algorithm, the full search algorithm, while having the highest accuracy, but a very high computational complexity. How to achieve high accuracy and low computational complexity is motion estimation technology problems to be solved. So far there have been many fast block matching algorithm, these algorithms accuracy and computational complexity to achieve a better balance. Wherein, based on wavelet transform block matching algorithm is the use of the wavelet sub-band correlation coefficients, only molecules with a unit for matching operation, to predict the motion vector of the other sub-band as a block matching method, it is relative to other fast search algorithm to further reduce the computational complexity. This paper presents an improved square - diamond search algorithm to achieve medical image sequences motion estimation. This improved square - diamond algorithm reduces the number of search points. We will be applied to the medical image sequences in wavelet domain motion estimation, and digital subtraction angiography image sequence (DSA) experiments. The results show that the improved wavelet square - diamond algorithm is better than other algorithms with high accuracy. These fast block matching algorithms are based on the assumption: matching function monotonously changes, ie matching function value as the search point with the advantages of increasing the distance between. It is not used in practical applications, these algorithms tend to fall into local optimal solution. Based on the mechanism of natural selection were widely used genetic algorithms to solve the global optimum. But the speed of evolution of the standard genetic algorithm determines that it can not be directly used for motion estimation technique. Quantum genetic algorithm is a genetic algorithm combined with quantum computing. Chromosome encoding algorithm uses qubits, quantum gates mutation to evolve populations, in order to control the quantum current optimal solution to large variations make it a high degree of probability models evolve to adapt to than the traditional evolutionary strategy has faster convergence speed and global optimization capability. In this paper, a center-biased characteristics of motion vectors and quantum evolutionary strategy, adding in a quantum genetic algorithm initial population, presents a quantum genetic algorithm based on improved block matching method. Experimental results show that the proposed algorithm than three-step method, high precision, and with high probability higher than diamond search algorithm accuracy.