Separation of image parts using 2-D parallel form recursive filters

Area of Science:

• Digital Image Processing
• Signal Processing
• Computer Vision

Background:

• Composite images often contain multiple objects that require separation for analysis.
• Existing methods may struggle with occlusions and noise, limiting their effectiveness.

Purpose of the Study:

• To introduce a novel technique for separating objects within composite digital images.
• To adapt and extend the 2-D Steiglitz-McBride method for image decomposition.

Main Methods:

• Application of a parallel form extension of the 2-D Steiglitz-McBride method.
• Utilizing the discrete cosine transform (DCT) of the composite image.
• Decomposing the image into a sum of filters, each corresponding to an object's DCT.

Main Results:

• Preliminary results demonstrate successful separation of two objects from a composite image.
• The algorithm shows robustness in scenarios involving object occlusion.
• The method performs well even in the presence of moderate noise levels.

Conclusions:

• The proposed parallel form extension of the Steiglitz-McBride method is effective for object separation in composite images.
• The technique offers a promising solution for image decomposition tasks, handling complex scenarios like occlusion and noise.