For a long theory paper in image/signal processing, you could try IEEE Transactions on Information Theory. The papers typically run about 20 pages, but this classic runs 45 pages. Note that the two-column format can condense a one-column draft by 25% or more.
Trans IT is a very well-regarded journal in my field, with regular contributions from math-y engineers, statisticians, and engineer-y mathematicians. To quote the aims and scope (emphasis mine):
The IEEE Transactions on Information Theory publishes papers concerned with the transmission, processing, and utilization of information. While the boundaries of acceptable subject matter are intentionally not sharply delimited, its scope currently includes Shannon theory, coding theory and techniques, data compression, sequences, signal processing, detection and estimation, pattern recognition, learning and inference, communications and communication networks, complexity and cryptography, and quantum information theory and coding. IEEE Transactions on Information Theory papers normally contain a strong conceptual and/or analytical contribution.
The "strong conceptual and/or analytical contribution" means theorems and proofs; don't submit here if your paper doesn't have these. I couldn't find any stated page limit, but if your work is unnecessarily long for its contribution, the reviewers will definitely complain.
Another option is the SIAM Journal on Imaging Science. There is no hard-and-fast page limit there, but the policies do state that "papers exceeding 30 journal pages, excluding the supplementary material, will be reviewed more closely to ensure that the excess is fully justified."
For both of these journals, expect a very rigorous review cycle. I suspect that, unless the results are earth-shattering, you will still hear complaints about the length.