Are pedersen hashes of small inputs safe?

I understand that the end result of a Pedersen Hash (like this one) is a point in an Elliptic Curve.

In the example implementation mentioned above, the input $M$ is split into chunks of 200 bits (the last one possibly being smaller). For each chunk, disconnected/random points in the Elliptic Curve are generated and the end result is a linear combination of those points, with the coefficients depending on the bits present in each chunk.

My question is: suppose I wanted to hash something 200 bits long. I would therefore only need one chunk and one generated point. Of course, this point would be multiplied by a scalar generated by the bits in the chunk to give the resulting hash. Would this be a “secure” hash? Or should I split $M$ into smaller chunks so as to have at least a minimum amount of different points to combine linearly?

Thanks!

A Pedersen hash only guarantees collision resistance. If you have a single base of prime order and an input smaller than that order, then the "hash" is one-to-one so it is perfectly collision-resistant.

However, using a hash in a non-compressing application is an almost-sure sign that your protocol relies on other/additional security properties to collision resistance. Depending on what those properties are, a Pedersen hash may not be appropriate at all for the application, regardless of how many bases are used.