Research On The Quantum-Inspired Method For Image Processing

The quantum-inspired method is a new method or improvement onthe method available now which is created by borrowing fundamentalconception and interesting principles from quantum theory for solvingsome practical problem. The quantum-inspired method for imageprocessing is an extension of quantum-inspired method to imageprocessing field. It provides a new idea for the theory research andtechnological realization in image processing field.In this dissertation, quantum throry is reguarded as an anxiomssyetem based the fundamental postulates of quantum mechanics andapplied to created and describe those quantum-inspired methods forimage processing proposed here. The concrete work of the dissertation issummarized as follows.An improvement on mathematics morphology is made here byapplying of the principles about superposition and measurement ofquantum state. A new structure element similar to a superposition ofstates in quantum mechanics, nemaed quantum superposition structureelement, short for QSE, is proposed and constrcted and a newmorphological operation mechanism based QSE, called collapsingmorphological operation is defined. The shape of QSE in collapsingmorphological operation can be varied with its moving location in imagethrough a measurement to QSE by an operator. A method used forproducing the operator used for measuring according to local feature ofimage is proposed and the methods use for self-adapted filtering and edgedetecting of image based collapsing morphological operation are createdfrom this. The result obtained in the computer simulation of thisself-adapted filtering show that this filtering method has more powerfulcapability to filter noise in image than customary morpgologyfilter.Therefore, the problem of how to select a suitable structure elementin morphology operation can be solved. The result obtained in thecomputer simulation of this edge detecting method show that this methodcan detect directly edge of image corrupted by noise without pre-filtering.Therefore, the damaging of edge in pre-filtering can be avoied. A self-adaped median filtering is also proposed in thisdissertation.The mask used for median operation is produced according todistribution of gray scale of pixels in the neighborhood of its movinglocation in image. In order to realize this, each gray scale of all pixels inthe neighborhood is transformed into quantum bit and is operatedsuccessively by quantum gate. Then, a random observation of the pixelsis carried to make them collapsing to produce a mask adapted for themoving location. This improved median filtering is a self-adaptedfiltering because of mask varied with local feature of image in its movinglocation. Therefore,the result obtained by applying this median filter isbetter than that obtained by conventional median filter. The principle ofthis median filterin is described in mathematics formulation and itsrealization simulated in computer is made here. The result obtained in thesimulation show that this method can not only efficiently filter the noisein image and reserve the detail of image fairly good, but also is notsensitive to the intensity of noise in image.Another important content in this dissertation is how to applyingquantum throry as an anxioms syetem to creat new edge detecting methodof image. In this new method created here, the relation between twopixels is described by means of the superposition of two quantum bitcorresponding to their gray scales. Quantum entanglement is used fordescribing the change of gray scale of the pixels and the average of linearhermitian operator in the superposition of two quantum bit is used forjudging the edge of image. A function of this new edge detecting methodto compensate visual sense of human is proved theoretically and isverified in computer simulation. This method agrees with the demand ofmachine's visual sense because the edge resulting from a very litterchange in gray scale in the region of image with higher or lower gray canbe detected. The fact that the method can be degenerated that methodbased gradient operation on some condition is given, and also theauxiliary proof about rationality of this method by means of probabilitytheory is given too here.To solve the problem about mask optimizing in no-linear imagefiltering based on mask operation, a scheme with artificial intelligence for image filtering is also proposed here. This scheme consists of twoprocedures, namely study-optimization procedure and filtering procedure.An optimized mask can be obtained in the study-optimization procedurethrough learning from a set of image samples prepared and then ano-linear filtering for the image needed to processing is carried out by themask obtained in filtering procedure. This study-optimization procedureis realized by means of quantum-inspired genetic algorithm. In therealization, the masks are described by chromosomes coded in quantumbit and a random observation of these chromosomes are taken to makethem collapsing to produce a set of probable solution for the masks, andthe population consisted of these chromosomes is updated by applyingquantum rotation gate operator. The result obtained in the computersimulation to a median-opening-closing filter of this scheme shows thatthis algorithm proposed here is practicable.