A new inversion technique is presented for the identification of plasma filaments in wide-angle visible camera data. Direct inversion of camera data onto a field aligned basis is a poorly conditioned problem which is overcome by breaking the analysis into a `psuedo-inversion’ step followed by a `point spread function correction’ step. Camera images are first prepossessed, applying background subtraction, noise reduction and sharpening enhancements to bring out the transient filaments structures in the images. A large collections of equilibrium magnetic field lines are traced and project onto the camera field of view. These field line images are combined to form a geometry matrix which is used to produce a pseudo-inversion which is obtained from a convolution of each individual field line image with the camera image. A tractable inversion is then performed on a point spread function matrix which is derived from the geometry matrix. The resulting 2D intensity distribution parameterised by the field line machine coordinates at the mid-plane of the machine is a good approximation of the direct inversion problem. Blobs of high intensity are identified using the watershed algorithm and 2D Gaussians are fitted to get the positions, widths and amplitudes of the filaments. A synthetic diagnostic producing artificial camera data containing experimentally representative filaments is utilised to rigorously benchmark the accuracy and reliability of the technique. 75% of input filaments above the detection amplitude threshold are successfully detected and 11% of detected filaments are found to be erroneous. The accuracy with which filament properties and their probability density functions are recovered is discussed, along with sources of error and methods to minimise them.