COMP322/S2000/L181 Pre-processing: Smooth a Binary Image After binarization of a grey level image, the resulting binary image may have zero’s (white) and one’s (black) at undesirable places. Examples: “Black Spot”“Hole” “Missing Corner” “Bump”
COMP322/S2000/L182 Pre-processing: Smooth a Binary Image Need to define algorithms to “remove” these “unwanted” one’s and zero’s. One possible solution: use a 3x3 template Image window centered at p : Basic Boolean Operations: Complement: AND OR+
COMP322/S2000/L183 Pre-processing: Smooth a Binary Image Boolean Expressions: (1- black; 0 - white) (Examples given in class) B1: B2: B3: B4:
COMP322/S2000/L184 Pre-processing: Smooth a Binary Image Boolean Expressions: (1- black; 0 - white) (Examples given in class) B5: B6:
COMP322/S2000/L185 Robot Vision System Major Phases in Robot Vision Systems: A. Data (image) acquisition l Illumination, i.e. lighting consideration l Lenses, and Cameras l Digitizers B. Pre-processing l Enhancement, i.e. smoothing, edge detection l Segmentation (Binarization) C. Recognition l Feature extraction l Pattern matching D. Part Manipulation l Choice of robot, gripper l Orientations of gripper, camera, part w.r.t robot base
COMP322/S2000/L186 Recognition: Feature Extraction Example: A binary image Question: How to identify these 2 objects? Objects are connected components ==> Connectivity ==> 4-connected (4-neighbours) ==> 8-connected (8-neighbours) Each object has different labels ==> Labeling
COMP322/S2000/L187 Recognition: Feature Extraction Labeling: Given a binary image, different objects has different labels. (no one perfect algorithm) Idea: - Proceed from top-left (0,0) to bottom right (N x, N y ) searching for “1”; Give the same label if 8-connected; Give a different label if not connected
COMP322/S2000/L188 Recognition: Feature Extraction Boundary Representation: (Chain code) Convention (8-connected): Example: Starting pixel is (2,1) Boundary: (0,0,0,7,5,4,4,4,3,1) (Try other examples on your own) (x,y)
COMP322/S2000/L189 Recognition: Region Representation Run Length Encoding: Idea is to reduce storage. The lengths of “1”s and “0”s are encoded. The first integer indicates the value of the starting pixel and is followed by the lengths Example: (image: 9x5 bytes) Run Length: (0,11,4,6,4,4,4,6,5,1) (10 bytes) Run Length: (1,2,2,3,4,3,8,2,4,3,8,3,3) (13 bytes)
COMP322/S2000/L1810 Recognition: Region Representation Quad-tree Representation: Idea:- Root represents the whole image - each node has maximum of 4 descendants - each node represents (2 i x 2 i ) pixels of the same value, (i=0,..., k). Examples given in class.