You cannot use imread() in HDL. HDL has no I/O system.
When you are building for HDL, you have to assign to each variable before it is used as the output of a call. The assignment will determine the size and data type of the variable, which will be important for determining how many memory cells to allocate in the HDL. The maximum size of variables must be strictly defined because there is no dynamic memory allocation in HDL.
You should be recoding your algorithm to take as input a vector of values of fixed size (or at least maximum fixed size if the sizes are passed in). You can then reshape() to array for processing. The vector form is necessary because you need to cut down to a small number of I/O pins because FPGA do not have large pin-outs. Remember, one pin for every bit for every value that is to be input simultaneously, so you want to have a narrow bus (at high speed) to transfer your image array.
