A piece of code that can extract shapes from bitmaps is needed

Hello I'm interesting your project very well I'm a Good C++/C#, Java, OpenCV, Math, Algorithm expert. I m quite well experienced in these jobs. Let's go ahead with me I want to service for you continously. Thanks

Hi, please answer these questions: how exactly will the shapes be defined? will the shapes always be of the same size? will they always be of the same color? will the bitmaps contain just the shapes or other things too (like background)? can you give some examples of real bitmaps you will be using this program with? Also I don't fully understand your example: two letters ("a" and "t") are missing in the bottom row, why? Also why does the third last red rectangle in the bottom row contain the end points of the previous letter "t"? Do you consider that as well extracted? The same is true about the last two shapes that are not properly separated either (the second last red rectangle contains a part of the last shape). When you answer these questions, I can tell you my approach.

I am an experienced programmer, with more than 10 years of experience. I understand exactly what you need. I promise you will be satisfied with my work. Thank you for your trust.

Hi, are your shapes always separated from the backgrund that easily ( like black on white, or like shape below, background above teh bitmap.s mean level or some other trivial way to determine whether pixel belongs to a shape or background.) Also are the shaes separated from each other, no touching and overlap, It this is the case s imple flood fill algortihm would solve this (keeping trask of min max x/y coordinates, I guess the bounding box is axis aligned). For grater speed and lower stack usae a line scan flood fill is better , 4 or 8 connectivity, I inded use such thignfor some object detecion on job programs .. only a little bit bad think i I am less C# expariaenced, but Iguess I can do this if I have some time to se some bitmap loading and such auxilary things...

I have something similar. I used it to solve simple captcha, it is not 100%, but on this simple captcha it works with about 80% efficiency. I have to define all the shapes that are to be recognized in advance, a set of all shapes to be extracted. Then using very fast bitwise AND, XOR and OR operations defined shapes get compared with the test image. In my case 20% false detections comes from distortions in the test image. For now it works with a 5 character captcha, single thread on an older i5 resulting in about 100 images pro second getting processed. Set/number of shapes, size, and the size of the bitmap play a significant role in performance. My bid is also not final, depends on how good my solution suit your needs, how much customization is needed...

Hi, I’ve had a good look at your project description and I’m very interested in providing a solution. I have 4+ hands on experience on developing quality web solutions as well as desktop standalone solutions. I used to develop systems using latest adaptive technology such as 1. .NET/ASP.NET - MVC4,MVC5 - Entity Framework - SQL - C# (for .NET solutions) 2. JQuery/Javascript/HTML5 - Angular.js/Bootstrap etc (for designing and front end implementations) I want to start work immediately and awaiting for your positive response on skp id "agilepksolutions" so I can show you some of my Quality Development Work as well as we can discuss project in details. Thank you.

Hi The best approach from a performance point of view would depend on the size of the input images. With small images I would (1) identify background color (2) Flood fill a fuzzy match to that color with a chosen color (3) Any pixels not of the chosen color are the shapes (4) Flood fill those pixels to determine a min/max on the X,Y planes to form a rectangle that will encapsulate it (5) repeat until all non chosen color pixels are gone. We then have a list of rectangles for the shapes. If it is the shapes you want and not the rectangles then the second flood fill can be eliminated. The original flood fill can store the pixels that it isn't changing into connected collections. On larger images, or noisy images, a D* implementation, or edge detection such as Canny, to find edges and a contour trace algorithm, possibly from the openCV library, would likely be more performant in finding the X,Y bounds. There will be a cut-off image size/noise ratio where one technique outperforms the other. Without knowing the reason for detecting the shapes it is difficult to be more specific. Your example image show irregularities in the boxing at the end of the sequence where the letters turn into shapes. Without knowing what you are trying to achieve, I am unsure why you have chosen to box those shapes as you have. Please feel free to ask questions if you have any Philip Smith

Hello. My solution is: 1)represent the picture as a graph: white pixels are nodes; pixels are adjacent if they are neighbors on the picture. 2)perform depth first search on that graph to find its connected components. 3)analyse each connected component in order to classify it as one of the shapes: shape is also a graph so if this two graphs are similar (it's the tricky moment - to invent similarity function) then mark the component as a shape.

I will approach the problem with BFS, achieving complexity of O(H*W), where H and W are respectively height and width of the bitmap. The program will return the set of shapes. The unclear part of the task however is the extraction of the characters. In the example, the letter 'i' is marked as two shapes, which runs contrary to your requirements, because it is not marked as a letter. If there is a mistake and you would like the program to recognize letters, please define what "well-defined" means. For example, whether the font is given or not.

Hi! My name is Konstantin. I'm an expert in real-time images and video processing, information extraction and objects recognition. Task you want to solve is relatively easy if you don't need to do it fast. If you need really quick and resource critical implementation number of special tricks can be used. It this particular case source image should be binarized, filtered from binary noise. Next binary 8-linked search should be used for shapes outlining with following image subdividing using found binary shape masks. All those steps can be really fast in case of proper implementation. Also it's possible to extract semi-overlapped shapes correctly, i.e. the last "t" letter and following "box-with-handle" shapes can be extracted with all belonging pixels separating them into bit bigger bounding rectangle. Feel free to ask any questions

- Experienced in Steganography algorithm which store plaintext into bitmap image - 5+ years experience with C#.NET - Will apply individual pixel scan in an image and then try to extract which pixel that isn't transparent or Alpha=0(Transparent pixel) and process it into a new bitmap picture based on each shape found.

I'd like to realize this in C++. I will enumerate closed shapes through flood filling, then compare with symbol templates