Optimal value (cvx_optval): +Inf
reports that the problem is primal infeasible, as CVX always reports relative to the problem as entered.
X*a == queryimage is overdetermined, and can not be solved exactly, but can for instance be solved by least squares, in which case the inf norm of the residual is 75.5. Or fit it in another norm, as below.
I don’t know what you are trying to do, but obviously your problem formulation is not “correct”.
Is this what you want?
n = size(X,2);
Optimal value (cvx_optval): +11125.7
Optimal value (cvx_optval): +11125.6
The least squares solution has a 1 norm of residual of 11909.5, which is a little bigger than the 1 norm of the residual of the 1 norm residual minimizer.