how process each region of a grayscale image
in [Matlab]
|
From: ImageAnalyst on 28 Mar 2010 23:30 On Mar 28, 9:44 pm, "Sami Oueslati" <Samy...(a)yahoo.fr> wrote: > Hope it works now:http://drop.io/mqlwmho# > > I mean process each region separately ------------------------------------------------------------- So just find the coordinates of each black bar's row and column. Then use those to crop out a small subregion and "treat" it as a complete image. What part is giving you trouble? Finding the bar locations? Cropping? "Treating"?
From: Walter Roberson on 29 Mar 2010 00:34 Sami Oueslati wrote: > the lines are just one pixel thick, for example the coordinates of > vetical lines are: > [17 37 58 82 100 120 145] > > and for horizontal are: [24 47 72 93 116 138] > > I need strictly parts inside, every rectangular region without lines to > measure the mean gray level of each one I have examined your image in fine detail, and your grid lines are NOT one pixel thick. You can see this clearly near 100, where there is a tick mark right at a grid line, and the tick mark is definitely thinner than the grid line. In the horizontal direction, the grid lines seem to be three pixels thick, and in the vertical direction, they might be four pixels thick (or perhaps it was just the one I was examining.) I cropped out the part of the image from the top left corner (0, 0) to (140, 140), which is the where the bottom-most and right-most tick marks are. The cropped image was 430 pixels high and 430 pixels wide. This is just slightly larger than a ratio of one coordinate unit rendering as three pixels (which would have given 420 x 420). The slight mismatch on the 3:1 ratio means that you will have to somehow detect the grid lines instead of just calculating where they should be in the image. I also examined the pixel values of the image and grid lines, and found that the grid lines are NOT a solid color. The grid lines are a range of colors from 0 up to about 19. However, some of the interior image pixels start from 12. If you select only pixels from 0 to 11 and make the rest white, then you will not have selected any of the interior pixels, but some of the grid line pixels would become white. If you set the threshhold any higher, so that the grid lines become solid, then you will also be selecting some of the real image pixels. If you look closely at the image pixels just above the bottom border at the 100 mark, you will see that there is a blur for a few pixels above the base. This and the not-quite-regular scale-factor suggests that this is not a mathematically constructed image, but is instead a scan taken of either a photograph or a printout, and that when it was converted to JPEG, the qualify factor was not set sufficiently high. JPEG gets worse and worse about sharp edges as the quality factor is reduced. This could also account for _some_ of the difference in pixel values for the grid lines, but examining the pixels in the 12 to 19 range supports the idea that there really is some overlap in the pixel values of the image and the grid. But then again, if it is indeed a scanned image, then inconsistent illumination could account for some of that.
From: Sami Oueslati on 26 Apr 2010 19:30 ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <def69c42-ddf9-4260-9ddd-a9427134edeb(a)19g2000yqu.googlegroups.com>... > On Mar 28, 9:44 pm, "Sami Oueslati" <Samy...(a)yahoo.fr> wrote: > > Hope it works now:http://drop.io/mqlwmho# > > > > I mean process each region separately > > ------------------------------------------------------------- > So just find the coordinates of each black bar's row and column. Then > use those to crop out a small subregion and "treat" it as a complete > image. What part is giving you trouble? Finding the bar locations? > Cropping? "Treating"? Hi, I've been absent for a long time, so I resume the subject... Here's the image: http://drop.io/mqlwmho# I have the coordinates of the black line that subdivide the image in subregions...My question is how to crop every subregion and how to extract the intensity of the center of each subregion???????????
From: ImageAnalyst on 26 Apr 2010 20:23 Just use regular indexing. You can find the rows and columns of the blanck lines dividing the subimages, can't you? Then do imageArray = fullSizeImage(row1:row2, col1:col2); meanGrayValue = mean(imageArray(:));
From: Sami Oueslati on 26 Apr 2010 20:37
ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <6bdc1204-a744-441d-a2c0-4a41e263c768(a)q8g2000vbm.googlegroups.com>... > Just use regular indexing. You can find the rows and columns of the > blanck lines dividing the subimages, can't you? Then do > imageArray = fullSizeImage(row1:row2, col1:col2); > meanGrayValue = mean(imageArray(:)); Sorry maybe you haven't understand me cause of my poor English... I know the coordinates of the black lines (I forgot the "s"), I obtained before and constructed the grid in the image...So my intention is to separate every subregion (of the grid) formed by the black lines...and by the way determine the intensity value of the center of each subregion. |