文件列表:

graphcut (0, 2014-06-27)
graphcut\SegTool (0, 2014-06-27)
graphcut\SegTool\CODE_MAP.pdf (31116, 2006-05-18)
graphcut\SegTool\SegToolBox (0, 2014-06-27)
graphcut\SegTool\SegToolBox\SegToolBox (0, 2014-06-27)
graphcut\SegTool\SegToolBox\SegToolBox\AutoCut.m (855, 2006-05-17)
graphcut\SegTool\SegToolBox\SegToolBox\AutoCutRefSeg.m (1636, 2006-05-16)
graphcut\SegTool\SegToolBox\SegToolBox\AutoCutRefine.m (866, 2006-05-16)
graphcut\SegTool\SegToolBox\SegToolBox\AutoCutSeg.m (726, 2006-05-16)
graphcut\SegTool\SegToolBox\SegToolBox\GraphCutSegment.cpp (11517, 2006-05-18)
graphcut\SegTool\SegToolBox\SegToolBox\GraphCutSegment.dll (45056, 2006-05-18)
graphcut\SegTool\SegToolBox\SegToolBox\GraphCutSegmentLazy.cpp (10532, 2006-05-16)
graphcut\SegTool\SegToolBox\SegToolBox\GraphCutSegmentLazy.dll (40960, 2006-05-18)
graphcut\SegTool\SegToolBox\SegToolBox\SegResult.mat (89336, 2014-06-25)
graphcut\SegTool\SegToolBox\SegToolBox\SegTool.m (1137, 2013-07-14)
graphcut\SegTool\SegToolBox\SegToolBox\Segment.m (3237, 2006-05-17)
graphcut\SegTool\SegToolBox\SegToolBox\SegmentBoth.m (9046, 2006-04-29)
graphcut\SegTool\SegToolBox\SegToolBox\SegmentGC.m (8578, 2006-05-17)
graphcut\SegTool\SegToolBox\SegToolBox\SegmentRefine.m (7789, 2006-05-17)
graphcut\SegTool\SegToolBox\SegToolBox\SmartSelect.m (909, 2006-05-16)
graphcut\SegTool\SegToolBox\SegToolBox\SmartSelectSeg.m (2669, 2013-07-14)
graphcut\SegTool\SegToolBox\SegToolBox\bginput.m (1327, 2006-05-17)
graphcut\SegTool\SegToolBox\SegToolBox\block.h (8191, 2006-03-05)
graphcut\SegTool\SegToolBox\SegToolBox\datain.m (762, 2006-05-16)
graphcut\SegTool\SegToolBox\SegToolBox\dataout.m (704, 2006-05-17)
graphcut\SegTool\SegToolBox\SegToolBox\fginput.m (1292, 2006-05-17)
graphcut\SegTool\SegToolBox\SegToolBox\graph.cpp (2277, 2006-03-05)
graphcut\SegTool\SegToolBox\SegToolBox\graph.h (5798, 2006-03-05)
graphcut\SegTool\SegToolBox\SegToolBox\helper_functions.h (1086, 2006-05-16)
graphcut\SegTool\SegToolBox\SegToolBox\images (0, 2014-06-27)
graphcut\SegTool\SegToolBox\SegToolBox\images\black_car.png (52656, 2004-11-10)
graphcut\SegTool\SegToolBox\SegToolBox\images\blue_car.png (80025, 2004-11-10)
graphcut\SegTool\SegToolBox\SegToolBox\images\carsten.jpg (40968, 2006-04-06)
graphcut\SegTool\SegToolBox\SegToolBox\images\fish.png (288024, 2006-05-01)
graphcut\SegTool\SegToolBox\SegToolBox\images\flowers.png (285543, 2006-05-01)
graphcut\SegTool\SegToolBox\SegToolBox\images\food.bmp (236250, 2006-03-19)
graphcut\SegTool\SegToolBox\SegToolBox\images\horse.jpg (33020, 2006-04-06)
graphcut\SegTool\SegToolBox\SegToolBox\images\llama.bmp (571394, 2006-03-28)
... ...

THIS FILE INCLUDES USAGE INSTRUCTIONS FOR SEGMENTATION TOOLBOX ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Authors - Mohit Gupta, Krishnan Ramnath Affiliation - Robotics Institute, CMU, Pittsburgh 2006-05-15 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We provide a freely available Segmentation Toolbox that can readily be used for image segmentation tasks. Please acknowledge the authors if you are using this ToolBox. ************************************************************* ************************************************************* Important Notes: 1) The parameters file 'mksopt.m' contains declarations for all the tunable parametes. 2) To segment a new image, please close the present working window and select a tool from the SegTool menu. Then proceed to select a new image from the file menu and perform operations on that image. 3) Modifications required to run the code in "Batch Mode" (for example, on PASCAL database): "Use the core segmentation code (Segment.m, (LZ) , SegmentGC (GC) ) for running the algorithm on huge data sets, without using the gui. You just needs to write a wrapper function that calls these functions repeatedly for each image in the dataset. 4) How are the outputs stored?: The outputs of each segmentation is stored in SegResult.mat file. This has the resulting segmentation mask (SegMask) and the segmented image (SegImage.) *************************************************************** *************************************************************** Please follow the instructions below for using the toolbox. Follow specific instructions for each part of the toolbox. -------------- STEP 1: -------------- The Segmentation ToolBox is invoked by typing "SegTool" at the Matlab commandline. -------------- STEP 2: -------------- The Segmentation ToolBox has three parts: 1) SmartSelect (Lazy Snapping) Tool : This tool is used to perform supervised segmentation by marking foreground and background seeds. 2) AutoCut (Grab CUt) Tool: This tool is used to perform semi-automatic segmentation by just marking a bounding rectangle (two opposite corners) covering the object to be segmented. 3) AutoCutRefine (Grab Cut + Lazy Snapping) Tool: This allows the user to perform a SmartSelect Refinement on the AutoCut result. -------------- STEP 3: -------------- There are separate instructions for each tool: %%%%%%%%%%%%%%%%%%%%%%%% 1) SMARTSELECT: %%%%%%%%%%%%%%%%%%%%%%%% - Choose SegTool->SmartSelect - This opens a file menu. From this choose the image you want to segment. - The image is displayed, along with two menu options on the right: 1) SmartRefine 2) SmartRectangle These indicate the two ways in which we can mark seeds on the image to initiate the graph cut segmentation algorithm. We can toggle between the two ways by choosing the corresponding radio button $$$$$$$$$$$$$$$$$$ 1) SMARTRECTANGLE: $$$$$$$$$$$$$$$$$$ Using this we can mark seeds in the form of rectangular regions on the image. Choose SmartRectangle - This gives you three options: 1) Foreground - Click this to mark foreground seeds 2) Background - Click this to mark background seeds 3) GraphCuts - Click this to run the segmentation. Once you click "foreground" or "background" the system waits for input. The input in this case is two corners of the rectangle (usually top left and bottom right) that the user selects. Clicking on foreground (or background) button and then choosing two points plots a rectangle which indicates the seeds. (Foreground rectangles are red, background rectangles are blue) To click a new rectangle the user has to click the foreground or the background button again. $$$$$$$$$$$$$$$ 2) SMARTREFINE: $$$$$$$$$$$$$$$ User can use this to specify seeds in the form of strokes. Choose SmartRefine radio button - This gives you three options: 1) Foreground - Click this to mark foreground seeds 2) Background - Click this to mark background seeds 3) GraphCuts - Click this to run the segmentation. Once you click "foreground" or "background" the system waits for input. The input is specified in the form of strokes. To start marking the stroke the user clicks on a point on the image and then moves the mouse. (The user should merely click, ^^^^^^NOT CLICK AND HOLD^^^^^ while moving.) The user marks the end of the stroke by a second click. The user can click as many strokes as he wants by again clicking once on an image point to initiate the stroke and then click once again to stop marking the stroke. Sometimes the strokes take time to plot, so if the system is busy the user may not see the strokes immediately. Once the user has chosen enough seeds he can start the graph cuts algorithm by clicking on graphcuts. This runs graphcuts and gives the result. ^^^THE USER CAN REFINE THIS RESULT^^^: by choosing the seeds by doing the same as described above on the same image. These seeds are added to the previous seeds when the user calls graphcuts again. %%%%%%%%%%%%%%%%%%%% 2) AUTOCUT: %%%%%%%%%%%%%%%%%%%% - Choose SegTool->AutoCut - This opens a file menu. From this choose the image you want to segment. - Once the image is chosen it is displayed and the system waits for user input. - The input is two points, which indicate the corners of a rectangle ^^^^^COVERING THE OBJECT TO BE SEGMENTED^^^^^^^. - Usually the user chooses the two points as the top left and bottom right corners of the rectangle over the object to be segmented. - The algorithm iterates and gives a final segmented result. %%%%%%%%%%%%%%%%%%%%% 3) AUTOCUTREFINE: %%%%%%%%%%%%%%%%%%%%% - Choose SegTool->AutoCutRefine - This opens a file menu. From this choose the image you want to segment. - Once the image is chosen it is displayed and the system waits for user input. - The input is two points, which indicate the corners of a rectangle ^^^^^COVERING THE OBJECT TO BE SEGMENTED^^^^^^^. - Usually the user chooses the two points as the top left and bottom right corners of the rectangle over the object to be segmented. - The algorithm iterates and gives a segmented result. - The segmented result is displayed as a mask image and also as a green boundary on the input image. - The user can now refine this segment by using the SegmentRefine tool (see SmartRefine above.) - The user accesses the refinement tool from the right menu. As before he can choose Foreground or Background seeds in the form of strokes (A click to start stroke, a click to stop stroke.) - As before he marks foreground and background strokes and then clicks on SegmentRefine to refine the segmentation. Remember: AUTOCUTREFINE = AUTOCUT + SMARTSELECT (SMARTREFINE) -----END------ Please refer to README_COMPILE for compilation instructions. Please refer to CODE_MAP for a list of functions and the call structure.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：