Finding a largest rectangular object in binary image with minimum length and width

How to find a largest rectangular object in binary image. Any connected object should be ignored. The object should have a minimum length of 50 pixels and minimum width of 4 pixels. The examples images are attached.

Thanks

[Image Analyst]

Not sure I understand. Each of the yellow shapes can be considered as a collection of many touching rectangles. And a huge number of them since they can be constructed in many different ways using differently shaped and sized rectangles. So should they all be ignored? You'd have nothing left.

Why do you say some of them have no rectangle in them.

In general to find the largest blob you can do

largestBlob = bwareafilt(binaryImage, 1);

but that works on all blobs, regardless if they're perfectly rectangular.

[Az]

The right rectangular should have a minimum size (mirnumum width and minimum length). The image on left has that size of rectangular

[Image Analyst]

According to your definition, how many rectangles are in each image and where are they? Because I see either no rectangle at all, or a shape made up of many, many rectangles. Can you outline their bounding boxes in red or green or something so I can see where they are? Also, can you attach each image individually with the paper clip icon?

[Walter Roberson]

The code I posted in my Answer searches for rectangles and throws away the ones that are too small, and finds the largest of what remains.

The remaining question is whether you did intend "width" to refer to horizontal extent, or if instead you are referring to "length" as the longer of (horizontal extent, vertical extent) and "width" as the smaller of the two ?

[Az]

Both regionprops(BoundingBox) as well as Bwareafilt results in selecting a biggest object.

I want to to find:

biggest rectangular region

Shorter side at least 3 (horizontal or vertical)

Longer side at least 50 (horizontal or vertical)

The region can also be a part of an object (the parts of the object not contributing to the rectangle should be ignored)

[Walter Roberson]

When you indicated earlier that "Any connected object should be ignored" my understanding was that if you had rectangles that were connected to any other shape, that the entire blob should be ignored . That is why I coded regionprops to compare the convex area to the image area: those two values are only equal for blobs that are rectangles.

Answer 1

[Walter Roberson]

P = regionprops(YourBinaryImage, 'Area', 'ConvexArea', 'BoundingBox');
PA = vertcat(P.Area);
PCA = vertcat(P.ConvexArea);
PBB = vertcat(P.BoundingBox);
PW = PBB(:,3);
PH = PBB(:,4);
mask = PA == PCA & PH >= 50 & PW >= 4;
selected_P = P(mask);
[largest_object_area, idx] = max([selected_P.Area]);
largest_BB = selected_P(idx).BoundingBox;

Note that here I am interpreting "width" as horizontal extent, and then by default that means "length" in your question must refer to vertical extent.