Can you describe the background of this handwritten ?
If it a constant background you can use the threshold method.
Also you can use the watersheld methods but you need a post step to separate the handwritten regions from background regions. You can do that by calculating the area of the regions then based on these calculation, an area threshold puls checking if the region centroids approximately in the same line can be used to separate the handwritten regions from background regions.
Apart from Mohamed's thresholding suggestion, you can also try other simple techniques such as k-means clustering, with about 3 clusters (1-background, 2-light markings, 3-heavy markings, i.e. if you are using the greyscale image) followed by Connected Component Analysis to obtain each connected character region separately. Two functions currently exist in matlab for performing these operations, kmeans and bwlabel.
can you post up a sample picture of the handwritten characters, and a brief description of the final result that you are trying to achieve. That way I would have a better understanding of the problem.
Here I attached the input image. I have done line segmentation and word segmentation. Now I am not able to segment the character. I solved few algorithms, but I didn't get prefect segmentation.And i have done all preprocessing steps slant correction,skew correction and normalization.
I'm interested in the segmentation of English character. May I know if you have figured out the way to do the segmentation? If yes, can you share it to me. I need it in my FYP project.
You must have a invariant property of every symbol you want to process. For example with the number of ends of the line, the number of intersections, the number of holes, the number of bays, the number of headland, etc