Microscopic image fusion method based on region growing

摘要

A microscopic image fusion method based on region growing judges the fuzzy degree of the microscopic image and determines the fuzzy seed block by evaluating the definition of every image block in the microscopic image. Then, the fuzzy region and clear region are exactly segmented by region growing and are marked. Finally, a clear and high-quality microscopic image, fused by a plurality of microscopic images, is obtained. Due to the combination of the definition evaluation of the microscopic image, and segmentation of the fuzzy region and clear region by region growing, the fusion results of the microscopic image of the present invention show great advantages at the subjective human perception and the objective evaluation. Furthermore, the present invention has simple calculation and stable result, is easy to be implemented and adapted for fusing the digital optical microscopic images which are shot by shallow depth of field.

主权项

1. A microscopic image fusion method based on region growing, comprising steps of:
(1) setting {IS1(x,y)} to one microscopic image collected under an optical microscope, and setting {IS2(x,y)} to another microscopic image collected under the optical microscope, wherein 0≦x≦W−1, 0≦y≦H−1, W denotes a width of an image, H denotes a height of the image, IS1(x,y) denotes a pixel value of a pixel point whose coordinate position is (x,y) in the {IS1(x,y)}, and IS2(x,y) denotes a pixel value of a pixel point whose coordinate position is (x,y) in the {IS2(x,y)}; (2) dividing the {IS1(x,y)} intoW×Hn×n  non-overlapping first image blocks, each of which having a size of n×n, recording a first image block with a coordinate position of (i,j) in the {IS1(x,y)} as IB1—n(i,j), dividing the {IS2(x,y)} intoW×Hn×n  non-overlapping second image blocks, each of which having a size of n×n, and recording a second image block with a coordinate position of (i,j) in the {IS2(x,y)} as IB2—n(i,j), wherein0≤i≤Wn-1,0≤j≤Hn-1,  and n is 2-power; (3) evaluating a definition of every first image block in the {IS1(x,y)} for obtaining a definition characteristic value corresponding to every first image block in the {IS1(x,y)}, and recording a corresponding definition characteristic value thereof as Fv1—n(i,j) for the first image block IB1—n(i,j) with the coordinate position of (i,j) in the {IS1(x,y)}; evaluating a definition of every second image block in the {IS2(x,y)} for obtaining a definition characteristic value corresponding to every second image block in the {IS2(x,y)}, and recording a corresponding definition characteristic value thereof as Fv2—n(i,j) for the second image block IB2—n(i,j) with the coordinate position of (i,j) in the {IS2(x,y)}; (4) determining a first decision threshold of the definition characteristic value corresponding to every first image block in the {IS1(x,y)}, recording the first decision threshold as Tn—1; and then judging whether every first image block in the {IS1(x,y)} is a fuzzy seed block according to the definition characteristic value corresponding to every first image block in the {IS1(x,y)} and the first decision threshold Tn—1, wherein for the first image block IB1—n(i,j) with the coordinate position of (i,j) in the {IS1(x,y)}, judge whether the definition characteristic value Fv1—n(i,j) corresponding to the IB1—n(i,j) is smaller than the first decision threshold Tn—1, if it's OK, then judge the IB1—n(i,j) as the fuzzy seed block and set a mark MB1—n(i,j) of every pixel point in the IB1—n(i,j) to 1, if not, judge the IB1—n(i,j) as the non-fuzzy seed block and set the mark MB1—n(i,j) of every pixel point in the IB1—n(i,j) to 0; and then according to the mark MB1—n(i,j) of every pixel point in every first image block in the {IS1(x,y)}, calculating a definition tagged image corresponding to the {IS1(x,y)} and recording the definition tagged image as {ID1—n(x,y)}, wherein ID1—n(x,y)=MB1—n(int(x/n),int(y/n)), ID1—n(x,y) denotes a mark of a pixel point with a coordinate position of (x,y) in the {ID1—n(x,y)}, and int( ) is a rounding operator; determining a second decision threshold of the definition characteristic value corresponding to every second image block in the {IS2(x,y)}, recording the second decision threshold as Tn—2; and then judging whether every second image block in the {IS2(x,y)} is a fuzzy seed block according to the definition characteristic value corresponding to every second image block in the {IS2(x,y)} and the second decision threshold Tn—2, wherein for the second image block IB2—n(i,j) with the coordinate position of (i,j) in the {IS2(x,y)}, judge whether the definition characteristic value Fv2—n(i,j) corresponding to the IB2—n(i,j) is smaller than the second decision threshold Tn—2, if it's OK, judge the IB2—n(i,j) as the fuzzy seed block and set a mark MB2—n(i,j) of every pixel point in the IB2—n(i,j) to 1, if not, judge the IB2—n(i,j) as the non-fuzzy seed block and set the mark MB2—n(i,j) Of every pixel point in the IB2—n(i,j) to 0; and then according to the mark MB2—n(i,j) Of every pixel point in every second image block in the {IS2(x,y)}, calculating a definition tagged image corresponding to the {IS2(x,y)} and recording the definition tagged image as {ID2—n(x,y)}, wherein ID2—n(x,y)=MB2—n(int(x/n),int(y/n)), ID2—n(x,y) denotes a mark of a pixel point with a coordinate position of (x,y) in the {ID2—n(x,y)}, and int( ) is a rounding operator; (5) setting n=128, n=64, n=32, n=16 and n=8, respectively repeating steps (3) to (4) for obtaining a definition tagged image {ID1—128(x,y)} corresponding to the {IS1(x,y)} and a definition tagged image {ID2—128(x,y)} corresponding to the {IS2(x,y)} while n=128; a definition tagged image {ID1—64(x,y)} corresponding to the {IS1(x,y)} and a definition tagged image {ID2—64(x,y)} corresponding to the {IS2(x,y)} while n=64; a definition tagged image {ID1—32(x,y)} corresponding to the {IS1(x,y)} and a definition tagged image {ID2—32(x,y)} corresponding to the {IS2(x,y)} while n=32; a definition tagged image {ID1—16(x,y)} corresponding to the {IS1(x,y)} and a definition tagged image {ID2—16(x,y)} corresponding to the {IS2(x,y)} while n=16; a definition tagged image {ID1—8(x,y)} corresponding to the {IS1(x,y)} and a definition tagged image {ID2—8(x,y)} corresponding to the {IS2(x,y)} while n=8; (6) based on the pixel points with the mark of 1 in the {ID1—128(x,y)}, {ID1—64(x,y)}, {ID1—32(x,y)}, {ID1—16(x,y)} and {ID1—8(x,y)}, making a fuzzy region growing for obtaining a region growing image corresponding to the {IS1(x,y)} and recording the region growing image as {IG1(x,y)}, wherein IG1(x,y) denotes a mark of a pixel point with a coordinate position of (x,y) in the {IG1(x,y)}; based on the pixel points with the mark of 1 in the {ID2—128(x,y)}, {ID2—64(x,y)}, {ID2—32(x,y)}, {ID2—16(x,y)} and {ID2—8(x,y)}, making a fuzzy region growing for obtaining a region growing image corresponding to the {IS2(x,y)} and recording the region growing image as IG2(x,y), wherein IG2(x,y) denotes a mark of a pixel point with a coordinate position of (x,y) in the {IG2(x,y)}; and (7) according to the region growing image {IG1(x,y)} corresponding to the {IS1(x,y)} and the region growing image {IG2(x,y)} corresponding to the {IS2(x,y)}, fusing the {IS1(x,y)} and the {IS2(x,y)} for obtaining a final fusion microscopic image and recording the final fusion microscopic image as {IF(x,y)}, whereinIF⁡(x,y)={IS⁢⁢1⁡(x,y),if⁢⁢IG⁢⁢1⁡(x,y)=0⁢⁢and⁢⁢IG⁢⁢2⁡(x,y)=1IS⁢⁢2⁡(x,y),if⁢⁢IG⁢⁢1⁡(x,y)=1⁢⁢and⁢⁢IG⁢⁢2⁡(x,y)=0Cif⁢⁢IG⁢⁢1⁡(x,y)=IG⁢⁢2⁡(x,y),  here, IF(x,y) denotes a pixel value of a pixel point with a coordinate position of (x,y) in the {IF(x,y)},C={IS⁢⁢1⁡(x,y),if⁢⁢Cnt⁢⁢1⁡(x,y)Cnt⁢⁢2⁡(x,y)IS⁢⁢1⁡(x,y)+IS⁢⁢2⁡(x,y)2,if⁢⁢Cnt⁢⁢1⁡(x,y)=Cnt⁢⁢2⁡(x,y),  Cnt1(x,y) denotes an amount of pixel points with a mark of 1 in an 8 neighbor of a pixel point with a coordinate position of (x,y) in the {IG1(x,y)}, and Cnt2(x,y) denotes an amount of pixel points with a mark of 1 in an 8 neighbor of a pixel point with a coordinate position of (x,y) in the {IG2(x,y)}.